Development of a Bayesian response-adaptive trial design for the Dexamethasone for Excessive Menstruation study

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Adaptive trial designs offer efficiency but require careful feature selection for feasibility.
• Bayesian response-adaptive designs dynamically adjust treatment allocation based on accumulating data.
• Optimizing dose-finding trials is crucial for effective and safe drug development.

Purpose of the Study:

• To identify key features of an efficient and feasible Bayesian response-adaptive dose-finding trial design.
• To evaluate Dexamethasone's efficacy in reducing excessive menstrual bleeding and determine optimal dosage.
• To simulate and compare various adaptive design options for maximizing learning about dose-response relationships.

Main Methods:

• A preparatory simulation study using a fractional factorial design was conducted.
• Virtual trial data were generated using SAS software for candidate adaptive designs.
• Bayesian model estimation was performed using WinBUGS, with results analyzed via Normal linear models.

Main Results:

• The simulation study evaluated the impact of design features (e.g., number of doses, adaptation rules) on trial efficiency.
• Empirical type I error and statistical power were estimated for different design scenarios.
• A final, robust adaptive design was identified through simulation, performing well across various conditions.

Conclusions:

• The study successfully identified a feasible and efficient Bayesian response-adaptive design through a simulation approach.
• The chosen design balances statistical rigor with practical implementation considerations for dose-finding trials.
• This methodology provides a framework for optimizing adaptive trial designs in pharmaceutical research.